Documentation

LeanPool.HighContrastHomogenization.Support.Book.Ch05.Theorems.Section54.OneStepContraction.GoodScaleInputs

Coarse-graining support: Support.Book.Ch05.Theorems.Section54.OneStepContraction.GoodScaleInputs #

Imported from the Apache-2.0 CoarseGraining development at commit c7ddd76c08ade64fed1b8d2ca51be14dfee8deb4.

Good-scale inputs for the one-step contraction #

This file repackages the public good-scale parameter bounds into the exact pieces consumed by the one-step contraction proof.

theorem HCPolySupport.Book.Ch05.Section54.OneStepContraction.goodScale_J_zero_le {d : ℕ} [NeZero d] {P : Ch04.RestrictionCoeffLaw d} (hP : Ch04.RestrictionLawCarrier P) (hStruct : Ch04.RestrictionStructuralLaw P) (hP4 : QuantitativeCoarseGrainedEllipticity P) {delta : ℝ} (hdelta_pos : 0 < delta) (hdelta_le : delta ≤ 1 / 2) {m : ℕ} (hgood_upper : hP.barSigmaAtScale hStruct 0 ≤ (1 + delta) * hP.barSigmaAtScale hStruct ↑m) (hgood_lower : (hP.barSigmaStarAtScale hStruct 0)⁻¹ ≤ (1 + delta) * (hP.barSigmaStarAtScale hStruct ↑m)⁻¹) (e : Vec d) (he : Ch02.vecNorm e = 1) :
have p_e := specialPAtScale hP hStruct (↑m) e; have q_e := specialQAtScale hP hStruct (↑m) e; Ch04.annealedResponseJAtScale P 0 p_e q_e ≤ (1 + delta) * √(thetaAtScale hP hStruct 0)

At a good scale, the unit-scale annealed response for the special vectors is controlled by sqrt(Theta_0).

theorem HCPolySupport.Book.Ch05.Section54.OneStepContraction.goodScale_tau_le {d : ℕ} [NeZero d] {P : Ch04.RestrictionCoeffLaw d} (hP : Ch04.RestrictionLawCarrier P) (hStruct : Ch04.RestrictionStructuralLaw P) (hP4 : QuantitativeCoarseGrainedEllipticity P) {delta : ℝ} (hdelta_pos : 0 < delta) (hdelta_le : delta ≤ 1 / 2) {m j : ℕ} (hj : j ≤ m) (hgood_upper : hP.barSigmaAtScale hStruct 0 ≤ (1 + delta) * hP.barSigmaAtScale hStruct ↑m) (hgood_lower : (hP.barSigmaStarAtScale hStruct 0)⁻¹ ≤ (1 + delta) * (hP.barSigmaStarAtScale hStruct ↑m)⁻¹) (e : Vec d) (he : Ch02.vecNorm e = 1) :
have p_e := specialPAtScale hP hStruct (↑m) e; have q_e := specialQAtScale hP hStruct (↑m) e; tauAtScale P (↑m) (↑j) p_e q_e ≤ delta * √(thetaAtScale hP hStruct 0)

At a good scale, all lower-scale additivity defects for the special vectors are controlled by delta * sqrt(Theta_0).

theorem HCPolySupport.Book.Ch05.Section54.OneStepContraction.goodScale_sigmaHat_inv_barSigma_zero_le {d : ℕ} [NeZero d] {P : Ch04.RestrictionCoeffLaw d} (hP : Ch04.RestrictionLawCarrier P) (hStruct : Ch04.RestrictionStructuralLaw P) (hP4 : QuantitativeCoarseGrainedEllipticity P) {delta : ℝ} (hdelta_pos : 0 < delta) (hdelta_le : delta ≤ 1 / 2) {m : ℕ} (hgood_upper : hP.barSigmaAtScale hStruct 0 ≤ (1 + delta) * hP.barSigmaAtScale hStruct ↑m) (hgood_lower : (hP.barSigmaStarAtScale hStruct 0)⁻¹ ≤ (1 + delta) * (hP.barSigmaStarAtScale hStruct ↑m)⁻¹) (e : Vec d) (he : Ch02.vecNorm e = 1) :
(sigmaHatAtScale hP hStruct ↑m)⁻¹ * hP.barSigmaAtScale hStruct 0 ≤ (1 + delta) * √(thetaAtScale hP hStruct 0)

Good-scale upper scalar-chain comparison in the normalized variables.

theorem HCPolySupport.Book.Ch05.Section54.OneStepContraction.goodScale_sigmaHat_barSigmaStar_zero_inv_le {d : ℕ} [NeZero d] {P : Ch04.RestrictionCoeffLaw d} (hP : Ch04.RestrictionLawCarrier P) (hStruct : Ch04.RestrictionStructuralLaw P) (hP4 : QuantitativeCoarseGrainedEllipticity P) {delta : ℝ} (hdelta_pos : 0 < delta) (hdelta_le : delta ≤ 1 / 2) {m : ℕ} (hgood_upper : hP.barSigmaAtScale hStruct 0 ≤ (1 + delta) * hP.barSigmaAtScale hStruct ↑m) (hgood_lower : (hP.barSigmaStarAtScale hStruct 0)⁻¹ ≤ (1 + delta) * (hP.barSigmaStarAtScale hStruct ↑m)⁻¹) (e : Vec d) (he : Ch02.vecNorm e = 1) :
sigmaHatAtScale hP hStruct ↑m * (hP.barSigmaStarAtScale hStruct 0)⁻¹ ≤ (1 + delta) * √(thetaAtScale hP hStruct 0)

Good-scale lower scalar-chain comparison in the normalized variables.

theorem HCPolySupport.Book.Ch05.Section54.OneStepContraction.goodScale_oneStepScalarWeight_le {d : ℕ} [NeZero d] {P : Ch04.RestrictionCoeffLaw d} (hP : Ch04.RestrictionLawCarrier P) (hStruct : Ch04.RestrictionStructuralLaw P) (hP4 : QuantitativeCoarseGrainedEllipticity P) {delta : ℝ} (hdelta_pos : 0 < delta) (hdelta_le : delta ≤ 1 / 2) {m : ℕ} (hgood_upper : hP.barSigmaAtScale hStruct 0 ≤ (1 + delta) * hP.barSigmaAtScale hStruct ↑m) (hgood_lower : (hP.barSigmaStarAtScale hStruct 0)⁻¹ ≤ (1 + delta) * (hP.barSigmaStarAtScale hStruct ↑m)⁻¹) (e : Vec d) (he : Ch02.vecNorm e = 1) :
oneStepScalarWeightAtScale hP hStruct m ≤ 3 * √(thetaAtScale hP hStruct 0)

The scalar weight multiplying the tau sum is bounded by a harmless constant times sqrt(Theta_0) at a good scale.